How AI and ML Are Used in HEOR
Data-driven health economics research can improve clinical and economic outcomes. Improvements like electronic medical records (EMRs) and other digital health solutions can help stakeholders make informed decisions and improve patient care in general.
A better understanding of how these cutting-edge technologies can improve healthcare and outcomes research is necessary.
How Can AI and Machine Learning be Used in Healthcare?
Health Economics and Outcomes Research (HEOR) is essential in the healthcare ecosystem to avoid potential resource losses via wasteful practices. HEOR plays an important role in reducing patient healthcare costs (e.g., assisting researchers and clinicians identify unnecessary care management procedures such as redundant tests that may result in delayed treatments for patients and diminish survival or remission rates for certain disease types.)
Healthcare providers and innovators are shifting from more traditional approaches to the application of ML and AI in this field. Today’s computing power and algorithms can process colossal amounts of data and, more importantly, provide more accuracy.
According to a McKinsey report, “big data could save pharma and medicine up to $100 billion dollars USD every year due to improved efficiencies in clinical trials, research and development, and decision-making with new tools that can help physicians, regulators, insurers, and consumers reign in costs and produce better patient health outcomes.”
The Future and Opportunities of Artificial Intelligence in HEOR
AI has enormous potential in HEOR. Challenges include improving the quality of healthcare information systems and data mining and storage, along with training decision makers and researchers on these new methods.
Some standard guidelines and methodologies for any AI-driven research will need to be established too.
Here are a few of the potential opportunities for using AI in HEOR:
- Natural language processing
- Text data analysis
- Machine learning
- Deep learning
The Potential of Machine and Deep Learning in HEOR
ML or statistical learning can be applied to HEOR to learn and perform tasks. Programs that embed ML learn from experience and augment the algorithm over time. The transformation of data to intelligent action in order to perform specific tasks can be beneficial in healthcare.
Another practical example is the use of neural networks in ML and algorithmic development. Neural networks can be used to update or adapt to a previously developed economic model. Furthermore, ML incorporates interoperability among processes and systems. The machine updates the results in real time for users to potentially use across healthcare systems of various regions, both nationally and internationally.
Deep learning algorithms can process and understand large amounts of data with hidden layers in a neural network; thus determining healthcare resource use, costs, and the impact of a patient’s given condition on health outcomes. This requires data from multiple sources, including data generated from individual patients, healthcare systems, and providers.
Deep learning techniques can be utilized to efficiently perform analyses using a combination of datasets from epidemiologic surveys, claims datasets, patient surveys, and registries.
AI, Privacy Issues, and Challenges
One of the biggest challenges of AI in recent years regarding clinical research is closely related to its use in other fields: privacy and data collection.
To enhance clinical practice, perform better cost-effectiveness analysis, and improve the population’s health, vast amounts of data need to be collected across various platforms. However, it is unclear how much of this information will be used in the future.
The collection of data and privacy issues is an even more sensitive matter when discussing patient medical and health records.
This data can be subject to similar security risks as those scrutinized in social media and other popular websites.
On top of that, there are still ethical concerns about AI and machine learning; specifically, when the technologies can be biased and favor particular subgroups over others—depending on the sources of information the algorithms are learning from.
Today, data quality is also in the spotlight because many processes pertinent to AI can be too complex to understand. More transparency is required to understand the methods, algorithms, and data types. There is still a lack of guidance in the reporting of models that use AI (ML specifically) in healthcare.
If AI is only centered on maximizing goals, it could lead to finding undesirable solutions, and we could encounter what is known as the control problem. In other words, how can we create machines that can help us without harming us?
Examples of AI Applications in HEOR
AI offers many healthcare applications and has seen solid growth in recent years. Here are some situations in which AI can help the healthcare field:
Disease Detection
Identifying patients at risk of having an undiagnosed or under-reported disease is a critical challenge. AI can help with it, and HEOR can quantify the burden of illness when an underdiagnosis occurs.
Disease Progression
AI can predict disease progression in patients across a range of medical conditions. AI programs may detect patients who may be at higher risk for rapid onset, especially for diseases with varied progression timelines. AI can assist practitioners with identifying and supporting patients who may benefit from more advanced treatment and/or closer monitoring. HEOR professionals can utilize the technology and analyze the cost of delayed treatment.
Hospital Readmissions
Predictive analytics can be utilized by healthcare professionals to aid patients with avoiding hospital readmissions for preventable conditions and thus improving overall care. The high cost of hospitalizations can then be reduced by evaluating the cost of such readmission.
Patient Adherence
AI can help identify key drivers of non-adherence in patient populations to guide targeted strategies for healthcare practitioners and researchers to understand why patients discontinue treatment. HEOR professionals can then analyze patient data and make conclusions on whether adherence leads to a reduction in healthcare costs and other economic outcomes, which will provide real-world evidence for subsequent adherence improvement strategies.
Adverse Events (AE) Management
AI can be used to identify patients at risk of experiencing an AE and the cost of the potential AE regarding HEOR. AI–based technology supports extraction from AE source documents and furthermore evaluation of case validity. It is viable to train the machine‐learning algorithms using safety database data fields as a surrogate for otherwise time‐consuming and costly direct annotation of source documents.
AI to Improve Healthcare and HEOR
AI will help the healthcare system process the increasingly growing amounts of information available in the field, resulting in better possible treatments and much more cost-efficient strategies.
However, advanced areas such as ML and deep learning are still experimental and must be studied more thoroughly, while privacy and bias issues will be an ever present concern to healthcare researchers, providers, and patients. The field is nascent, however, AI will continue to help improve healthcare services, research, and HEOR in the years to come.